A primary figure of merit for the performance of an IR photodetector is its normalized signal-to-noise ratio, or detectivity D*. When the detectivity is dominated by the dark electrical noise, the important functional dependence isD*∝QE/λJb1/2,  (1)where QE is the device external quantum efficiency, representing the fraction of incident photons that produce an electron-hole pair that is collected to produce electrical current, λ is the wavelength, and Jb is the dark current density (in the absence of any optical signal or background) at the operating bias. If the operating bias is zero, the dark current density is replaced by kT/R0A, where R0A is the resistance-area product at zero bias. The proportionality relation in Equation (1) also depends on some fundamental constants that are not affected by the device design.
Several previous works have discussed and simulated the potential advantages of resonant cavity infrared detectors (RCIDs). See J. G. A. Wehner et al., “Resonant Cavity-Enhanced Mercury Cadmium Telluride Detectors,” Journal of ELECTRONIC MATERIALS, Vol. 33, No. 6, 2004, pp. 604-608; see also L. Jun et al., “Design of a resonant-cavity-enhanced GaInAsSb/GaSb photodetector,” Sem. Sci. Technol. 19, 690 (2004).
RCIDs typically form a resonant cavity along the vertical axis by positioning two mirrors above and below the absorber. Thanks to the mirrors, any incident light with a wavelength tuned to the resonant mode of the cavity makes multiple passes through the absorber. This can allow a thin absorber positioned near the antinode of the cavity electric field to absorb most of the incident light for high QE, even when the absorber is much thinner than the absorption length without enhancement (1/α0).
The block schematic in FIG. 1 illustrates aspects of a generic resonant-cavity detector structure according to the prior art, configured for illumination from the top side thereof by light source 110. As can be seen in FIG. 1, such a structure in accordance with the prior art includes a bottom contact layer 102 on a bottom surface of an n-type substrate 101 (or otherwise positioned below the absorber region 106), a semiconductor bottom mirror 103, an n-type region 104 and p-type region 105 (in alternative configurations the p-type region may be positioned below the absorber sand the n-type region above) with a thin absorber region 106 between the n- and p-type regions, and a dielectric top mirror 108 (which may alternatively be a second semiconductor mirror). The top and bottom mirrors 103/108 form a resonant cavity 100 that significantly enhances the net absorption for high QE even when the absorber is much thinner than the 2-10 μm required to achieve high QE in a conventional broadband detector that does not employ a resonant cavity.
In other configurations known to the art, illumination is from the substrate side. In such cases, the “bottom” contact may be formed by patterning an annular ring on the substrate side of the structure (possibly after the substrate has been removed by polishing and/or etching), and the contact metallization deposited on top of the mesa may form part of the “top” mirror.
The detector mesa (which is typically circular or square) is etched to below the active absorber region 106, and in configurations employing illumination from the top an annular top metal contact layer 109 is deposited around its perimeter. Before metallization of the top contact, the mesa sidewalls and exposed region outside the mesa and below the junction (formed by the mesa etch) are coated with a dielectric 107 such as SiN to prevent shorting of the junction.
The QE at the cavity's resonance wavelength λres is given by:
                              QE          ⁡                      (                          λ                              λ                ⁢                                                                  ⁢                res                                      )                          =                              η            c                    ⁢                                                    SA                ⁡                                  (                                      1                    -                                          R                      1                                                        )                                            ⁢                              (                                  1                  +                                      R                    2                                                  )                                                                    (                                  1                  -                                                                                    R                        1                                            ⁢                                              R                        2                                                                                            )                            2                                                          (        2        )            
where R1 and R2 are the reflectivities of the top and bottom mirrors, respectively, S is the standing-wave enhancement factor that ranges from 0 to 2, depending on the position of the thin absorber with respect to the antinode of the cavity electric field, ηc is the carrier collection efficiency, which is expected to be nearly unity for the thin absorber, and A is the absorbance per pass in the absence of the cavity. The absorbance A of a thin absorber generally has a limit of A=α0d for a bulk-like absorber that does not reside within a resonant cavity, where α0(λ) is the absorption coefficient, and A=Mα for an absorber comprising one or more quantum wells (QWs), where M is the number of QWs and a is a dimensionless fraction representing the absorbance per pass by a single QW.
To maximize the photon absorbance within a relatively small spectral range near the resonance peak, the reflectivity R2 of the bottom mirror should be as close to unity as possible, while the reflectivity R1 of the top mirror can be varied to obtain the desired tradeoff between spectral width and absorption enhancement.
When R1 and R2 are close to unity, the QE can be high even when the absorber is extremely thin, e.g., a single QW. Another consequence of the cavity's high finesse is that the spectral linewidth over which the absorption is strongly enhanced narrows correspondingly. The full width at half maximum (FWHM) of the QE spectral peak (in wavelength units) is:
                    δλ        =                                                            (                                  1                  -                                                                                    R                        1                                            ⁢                                              R                        2                                                                                            )                            2                                                      π                ⁡                                  (                                                            R                      1                                        ⁢                                          R                      2                                                        )                                                            1                /                4                                              ⁢          FSR                                    (        3        )            
where FSR=λres2/(2 nL) is the free spectral range of the cavity, and n is the refractive index. The effective cavity length L includes both the thickness of the region between the mirrors and the penetration depths into the mirrors, which are non-negligible if the mirrors are realized as quarter-wavelength stacks.
If we further assume perfect collection of the photogenerated carriers (ηc=1) and that the thin absorber is positioned exactly at the resonant cavity's antinode (S=2), the on-resonance QE from Equation (2) and the FWHM of the spectral peak from Equation (3) become:
                              QE          ⁢                      (                          λ              res                        )                          =                                            4              ⁢                              A                ⁡                                  (                                      1                    -                                          R                      1                                                        )                                                                                    (                                  1                  -                                                            R                      1                                                                      )                            2                                ⁢                                          ⁢          and                                    (        4        )                                          δλ                      λ            res                          =                                                            (                                  1                  -                                                            R                      1                                                                      )                            2                                      π              ⁢                                                          ⁢                              R                1                                  1                  /                  4                                                              ⁢                                    λ              res                                      2              ⁢              nL                                                          (        5        )            
For broadband applications such as most instances of thermal imaging, it is disadvantageous to incorporate the detector's absorber region into a resonant cavity, since the loss of signal resulting from the much narrower spectral response more than offsets the absorbance enhancement provided by the resonant cavity. However, if the signal one wishes to detect is already narrow, the employment of a resonant cavity centered on the wavelength of interest can substantially increase the detectivity D* while retaining high QE, since the dark current is minimized by the very thin absorber region. A classic example for which the RCID configuration may be advantageous is laser-based chemical sensing of a trace gas, whose unique and narrow (<<0.1 nm) infrared absorption lines provide a fingerprint for identifying the given species and quantifying its concentration.
Challenges can arise when the resonant cavity approach is extended to longer wavelengths in the shortwave infrared (SWIR), midwave infrared (MWIR), and longwave infrared (LWIR) regions. The goal is to substantially enhance the detectivity, within a narrow spectral band, over that attainable using a state-of-the-art conventional broadband IR detector. While several RCIDs operating in the MWIR have been demonstrated previously (see below), to date none has exhibited a performance level competitive with that of a state-of-the-art conventional broadband IR detector operating at the same temperature and wavelength.
To our knowledge, all of the previous attempts to exploit a resonant-cavity absorption enhancement at a wavelength beyond 2.5 μm have performed at levels well below the state-of-the-art for conventional broadband IR detectors. Most of the previous RCID demonstrations have employed the lead-salt material system. See, e.g., M. Arnold et al., “Lead salt mid-IR photodetectors with narrow linewidth,” J. Cryst. Growth 278 (2005) 739-742; F. Felder et al., “Tunable lead-chalcogenide on Si resonant cavity enhanced midinfrared detector,” Appl. Phys. Lett. 91, 101102 (2007) (“Felder 2007”); F. Felder et al., “Lead Salt Resonant Cavity Enhanced Detector with MEMS Mirror,” Phys Proc. 3 (2010) 1127-1131 (“Felder 2010”); J. Wang et al., “Resonant-cavity-enhanced mid-infrared photodetector on a silicon platform,” Optics Express, Vol. 18, No. 12, 12890-12896 (2010); and J. Wang et al., “Monolithically integrated, resonant-cavity-enhanced dual-band mid-infrared photodetector on silicon,” Appl. Phys. Lett. 100, 211106 (2012).
However, conventional lead-salt IR detector materials generally suffer from short Shockley-Read lifetimes and high unintentional background doping levels, as compared to state-of-the-art HgCdTe and III-V IR detector materials. It is therefore not surprising that resonant-cavity lead-salt detectors are limited by similar materials-related issues. HgCdTe-based resonant-cavity designs have been proposed, see J. G. A. Werner et al., “Resonant Cavity-Enhanced Mercury Cadmium Telluride Detectors,” J. Electron. Mat. 33, 604 (2004), but to our knowledge have not been put into practice. Moreover, a monolithic semiconductor mirror technology suitable for a HgCdTe RCID has yet to be developed.
The III-V demonstrations to date have employed thick bulk absorber regions (typically d≈1 μm), which precluded significant enhancement of the detectivity. See Y. Shi et al., “Resonant Cavity Enhanced Heterojunction Phototransistors Based on GaInAsSb—AlGaAsSb Grown by Molecular Beam Epitaxy,” IEEE Phot. Tech. Lett., Vol. 10, No. 2, 258-260 (1998); A. M. Green et al., λ≈3 μm InAs resonant-cavity-enhanced photodetector,” Semicond. Sci. Technol. 18 (2003) 964-967; and A. M. Green et al., “Resonant-cavity-enhanced photodetectors and LEDs in the mid-infrared,” Physica E 20 (2004) 531-535. Since they also employed cavities with relatively low front-mirror reflectivity (R1<<1), those devices should be viewed as proof-of-concept demonstrations that were never intended to advance state-of-the-art performance.
In addition, some enhancement may, in fact, be achievable using conventional III-V bulk (e.g., InAs and InAsSb) or superlattice (e.g., InAs—Ga(In)Sb and InAs—InAsSb) absorbers in p-n junction or barrier (nBn and pBp) configurations from the prior art. See P. Martyniuk et al., “New concepts in infrared photodetector designs,” Appl. Phys. Rev. 1, 041102 (2014).
Because the resonant cavity configuration provides a strong enhancement of the net absorption over that resulting from a single pass of the light through the absorber, the absorber thickness can be shrunk to as little as ≈10 nm (e.g., a single QW) without sacrificing QE at the resonant wavelength. Therefore, because the dark current is reduced proportionally, the resonant cavity configuration will either provide higher detectivity D* at a given operating temperature, or maintain a target D* at higher operating temperature than is attainable using a conventional broadband IR detector.